INET Presentation

Ernesto Carrella

November 26, 2015

Fisheries today

Worm et al(2003): fishless seas by 2050
About 30% of the world fisheries have collapsed
The need for policy

Code example


    @Override
    public PerTripIterativeDestinationStrategy apply(FishState state) {

        MersenneTwisterFast random = state.random;
        NauticalMap map = state.getMap();


        final DefaultBeamHillClimbing algorithm = new DefaultBeamHillClimbing(
                alwaysCopyBest,
                dynamicFriendshipNetwork,
                stepSize.apply(random).intValue(),10);
        return new PerTripIterativeDestinationStrategy(
                new FavoriteDestinationStrategy(map, random), algorithm,
                probability.apply(state));
        }

A brief history in failure

Failures of the Commons and Gordon, 1954
Effort control and capital stuffing
Gear regulations and efficiency drops
Area closures and fishing the lines
Quotas and its allocation

Objectives

Policy Simulator
- Agents Flexibility
- Model Flexibility
Policy Optimization

The State of the Art

Random Utility Models
- Statistically Efficient
- Easily Generalizable
- Policy-Brittle
Dynamic Programming
- Strongly Rational
- Computationally Expensive
- Ad hoc

The One Agent Problem

Find the most profitable spot to fish
Constraints:
- No biomass information
- Environment changes over time
Subproblems:
- How to explore
- Explore-Exploit Tradeoff

One Agent world

One Agent sample run

Fast and Frugal Adaptation

How to explore
- Tow at a nearby cell from subjective best
- Random Hill-Climbing
When to explore
- Stochastically choose to explore next trip with probability \(p\)
- Adjust \(p\) if exploration is often (un)succesful
Why not more nuanced bandit algorithms?
Why not interview grounded algorithms?

Multiple Agents

Other boats consume biomass
You can use other boats information
How to imitate?
With probability \(p\) explore, with probability \(i\) ask a friend otherwise exploit